Transcription
So look, 11 months ago, we supported in a B2B service company acquisitions with decent revenue, terrible margins. The founder was burnt out, half of the team was coasting. So it's a classic turnaround setup. It's the kind of deal where most buyers bring in a new GM, hire three managers, and they hope for the best. We didn't hire anyone. What we did is we installed AI.
So today, that company runs on 30 or so AI agents and three humans only. Revenues up more than 40%. EBITDA went up by over 25%. The founder stayed on part-time because he actually enjoys the work right now. That's because the admin work that he hated just gone. And here's the part that surprises people. AI didn't just run the company after we bought it. AI found the deal, analyzed the financials, structured the offer, and helped us raise the capital to close it.
So, in this video, I'm going to show you the full life cycle. Basically, 10 different categories of AI agents. I'll share with you every prompt, every different agent in each department, the exact system running inside our acquisitions.com company as well, which is the same company that was supported in more than a billion dollar worth of transactions and supported me with selling rollups.com to Naval Ravikant's company. So, I built a master prompt that kick off this category. If you want, we probably going to put links below and you can join our workshop, our next buy AI and sell workshop where I show you how to install this live on real deals. So for today, let's start with what happens after you close because that's where the money is made. Then I'll show you how AI find and fund the deal in the first place.
So let's start with quick context about what we're doing here, which is understanding that after you acquire the company, you usually inherit some kind of a mess. In small businesses, every business has a bit of a mess. So here there are five different tools that no one documented in this company that we were part of. Processes that live on one person's head. I'm talking about reports that take sometimes a week to pull and you're paying for all of it, right? I'm talking salaries, benefits, software, office space while trying to figure out what the business actually does after you bought it. Most people try to fix those things by hiring more people. I try to fix things with AI. And I'm not talking about the chatbot that you have when you talk to ChatGPT because chatbot right now is kind of like just a toy, just a friend to talk to. You cannot put, I don't know, like a general ledger into ChatGPT and expect it to reconcile against your bank statements or flag expense anomalies or drop briefing into your inbox every day. It doesn't have access to your files. It doesn't remember what you told it yesterday. It cannot do things on repeat.
What we're using is cloud code. Cloud code is different. It lives on your machine. It reads your files. It writes outputs. It runs multiple step workflows. It checks also its own work and improves. And it saves everything locally or in the cloud. You can give it the full objective and goals. You can tell it, hey, process these invoices, flag anything over budget, update the AP aging, agent reports, and then you just walk away. The job becomes reviewing the output and being like the orchestrator of all these machines, not managing the actual processes yourself. So I'm sharing with you right now like a simple example. This is a kind of thing that you can run on day one. It takes a messy folder of business documents and maps out everything the company actually does. So, I'm going to put the prompt here. I won't bore you with the full prompt, but it says, "I just acquired a new company and I need you to help me understand what this business actually does operationally." Here's the situation. I have shared drive dump and then I have zero institutional knowledge. So, here are all the things you can put in the prompt, right? You're telling it to scan every file in this folder, document title, folder, key entities mentioned, you're going to put it there as well. The date range covered and then you tell it, hey, build me a business operations map in here. Here's the section one. It's the revenue, the delivery, the people, the systems, the vendors, the risks. Next thing is flag any gaps. Number four is output a one-page executive summary and start with the scan. This is what the output can look like.
Then when we move to the sales agent, sales paid for itself in 72 hours. I'm not exaggerating and here's why. The company that we were buying had one sales rep doing everything by hand. I'm talking prospecting, qualifying, writing proposals, following up. It's a classic founder sales that stopped scaling a few years ago. So what we did is we didn't hire more reps. I gave the one rep basically three agents to help with. In our case, agent number one was kind of like a signal agent. So it's at this point you need to forget cold outreach. But because this agent watches our entire target list and waits for something to change. A company raises funding that can be a trigger. They post a job for I don't know, head of marketing or M&A that means that the board just approved some kind of a strategy. Their competitor gets acquired, you can read that that means that they're going to be panicking or going to be the next acquisition. So this agent can catch up those different signals across the online platform. It can score them and every morning at 8:00 AM, there's a digest waiting with 10 to 15 warm accounts and personalized opening line for each one with the evidence URL and the data tabs. So here's the exact prompt and this is a real one. This is what we use, right? So it says basically I needed to build a signal-based outbound sales agent. You can see the full prompt. It says read the account list here. Every morning scan each account for each signal found this and return this. Generate personalized open line etc etc. You can see that and this is what the output can look like. Cold outbound in general gets you let's say 1 to 3% reply rate. This approach gets us 8 to 15%. Not because the emails are better, it's because we only contact people who are already thinking about the problem because we find their signals.
Agent number two in this department is what we call a CRM enrichment agent. So every lead that comes in gets auto-enriched. I'm talking about enriching with company data, text, employee count, recent news, competitive situation, etc., etc. Your sales team stops researching and starts selling the product. Agent number three in this department is the proposal generator. It reads the enriched lead data that we have. It generates custom proposals that reference the prospect's specific situation. And this is how we went from two hours per proposal to like 10 minutes of just reviewing everything. So three agents, one salesperson now that does the work or four because he's managing and orchestrating those different agents in his department.
Let's move to marketing. Marketing was the messiest department to automate. And honestly, the one where I see most people fail with AI. They use it to write blog posts. That sounds like every other blog post and sounds like AI. So, it's not what we did. What we did, we had four different agents here. First one is the landing page optimizer. The company we acquired had a landing page converting at like 1.2%. This AI agent read all the customer research, actual quotes from customers about what they care about. And it built a page around their language, not our language, their language. Basically, from where we were at, which is 1.2% to 3.8% in three weeks. Three builds. I didn't touch a single line of copy.
Second agent is the SEO and GEO engine. So, this is one most people aren't thinking about yet. SEO still matters, which is search engine optimization. Google still sends traffic. But here's the shift. Organic traffic to B2B sites especially is down 30% 30 to 50% year-over-year. That traffic moved into like AI into ChatGPT, cloud, perplexity. So those AI tools now intercept the click before it reaches your page. So what we do is we run SEO and GEO in parallel. SEO for Google, GEO for generative engine optimization. So basically for any AI situations, we create content specifically designed to be visited by AI tools when buyers ask about our category.
Third agent in this category is the content agent. So LinkedIn is where we can find like operators, decision-makers, depends on the business that you might have. The agent in our case watches the industry sources besides what type of post each input deserves. You write in the founder's voice, it generates variants. So before anything gets published, it gets scored. If the hook scores below 60 out of 100 potential engagement prediction, it gets rewritten or killed. So there's no more like posting and hoping. And a fourth agent that we did for this business is the Reddit engagement agent. This one is subtle and for most businesses you don't need it, but it's powerful for this specific business because 57 million people open Reddit every single day and a meaningful chunk of them are publicly asking questions that are product answers. So the agent monitors relevant subreddits, scores threads on buying intent, drafts helpful replies, and I stress helpful, not promotional, right? And it sends me daily digest. I copy, I paste, I post. It takes 10 minutes. It generates more qualified inbound for the business than paid ads. So, here's the content engine prompt that you can see. This is the one that generates like a full week of LinkedIn content scored for engagement before we even look at it. And that's kind of what you can expect here as well as far as output and whatnot.
Let's move to the next department. Everything related to operational agents. Operations is where the margin expansion actually happens. Sales grow the top line, ops grow the bottom line and post-acquisition the bottom line is what you really buy and care about. Right? So we have three agents here. The first one is process automation agent. And our rule here and this is non-negotiable across every company we're involved in acquisition. If any one of the team does something more than twice, it becomes a cloud call skill. I'm talking about invoice processing, vendor onboarding, compliance checks, client intake, even how we respond to support tickets. You don't write an SOP and hope someone follows it. You document their workflow once. It documents every step, every decision point, every edge case and packages it basically as a skill that anyone on the team can run with a single command after the first time. So let me show you what it looks like. This is the actual prompt we use when we try to turn manual processes into a skill and we get some amazing amazing outputs from that.
Secondly, we also have a project management agent. You can read our project tools summarize where everything stands, surface blockers before they become problems and your morning briefing in like you can get a 30-second briefing and get a lot of the knowledge in there as well. And the third thing we have is the quality control agent. That's one is super underrated in my opinion. It audits output against our documented SOPs. If someone deviates from the process, the agent catches it. Before we used to have like a manager for that business reviewing work. Now the manager focuses on strategy and the agent handles the quality and the delivery. So these three agents alone took EBITDA margins from around, I think 12 to above 20 like in the first 60 to 90 days.
Let's quickly go through the next ones which is finance agent. Finance is where most people are scared to let AI touch right now. I get it, numbers matter. Mistakes cost a lot of money. But the thing here is that AI doesn't make the decisions. It just surfaces the data for us so we can make better decisions. So here we have three agents. First one is the monthly close assistant. It can read the ledger, categorize transactions, flag anomalies, reconcile against bank statements. So you get month-end close went from like 5 days to one day in our case. It's not because AI is faster at math. It's because it just doesn't get distracted, right? It doesn't forget a line item. It doesn't run numbers to make them look good.
Then we also have the KPI one. The KPI one replaced our fractional CFO. So every Monday at 7:00 AM, we get a financial briefing lands in the prompt in the basically the project folder. We can see the cash position, the revenue by segment, the margin trends, the expense anomalies, the variance analysis, all in super plain English. It doesn't just say, hey, revenue went up. It says revenue went up 12%, but CAC went up 23%. So, we can really look at margins and really detailed things that in the past was like, I felt it was like annoying to ask your CFO for questions all the time or your accountant, whoever you're dealing with. Here's the full prompt and output that you can see here as well.
Third, we have the cash flow forecaster. It's a 13-week in our case rolling forecast that we like to see. It can read accounts receivables, aging, AP schedule, recurring expenses, seasonal patterns. It tells me exactly when cash gets tight and what to do about it. We used to pay a fractional CFO like a few thousand around $8,000 a month for exactly this type of work. This AI does it for like basically your cloud code subscription, $200 a month. Doesn't take a vacation in August, but most people do.
Next, we have HR agent. HR was a super cool win for us. In our case, I didn't expect AI to be that good there, but it turned out to be one of the highest ROI categories. We have two agents here. It's the hiring pipeline agent. It writes job descriptions from actual company context, not just generic stuff. So, it sounds like us. It attracts people that we like for the culture that we like. We got like 3x more qualified applicants because of the description and the follow-through to sending and posting and creating ads for recruitment work so well. We also have an onboarding agent that builds a custom onboarding plan for every new hire based on their role. So we have like day one tasks, week one goals, who to meet in the onboarding, what to read, check-ins. It's like super super cool. It's tracking every new hire and making them more productive and up to speed to get started with the actual work. I'll share with you some of the prompts that we use for this one and the output that we were able to see with that.
Next segment is the executive assistant plus KPI agent. So it's the last two categories and honestly these are the ones I personally use every single day. We have like an AI chief of staff. Every morning it generates a briefing on what happened yesterday across all the company, what's critical today, what needs my decision, give me like meeting prep for every call. It pulls relevant context from all of our systems. So I can walk into meetings knowing all the things about the people I'm going to meet. It triages my email, my Slack, it drafts responses. It saves me a lot of time and it's very, very cool. It's basically replacing almost like a full executive assistant that usually you need to pay between $50,000 to $100,000 a year.
Last agent we have is the KPI and exit readiness monitor. This one is super unique to what we do. It continuously scores the company, especially in our case like we have it for portfolio on exit readiness. You can see the revenue quality, the customer concentration, the margin stability, the growth trajectory. It tells me your estimated exit multiple today is let's say 5.8. Last month it was 5.2. And to get to 7x multiple, fix those three things. That's the sell part of the buy AI and sell process. You're not just running the business. You're running towards an exit. So, here's what it looks like as far as the prompt. This is what kicks off my day across different companies and portfolios.
So, everything I just showed you happens after you close. That's the AI install part. I want to quickly walk you through what it looks like because the same AI process that runs the company also found us the deal, analyzed the financials, structured the offer, helped raise the capital. So, there are a few more categories. I'll be real, real fast. We have the deal sourcing and screening. Like in our case, before we found that service company that we're talking about, right? We were doing sourcing the old way. I'm talking about like an analyst or like an employee to do outreach for brokers, emails like six hours a day just to find like 20 listings. Now with AI, we can do it at scale. We have agents that do that. We can filter things through like the revenue, the EBITDA, the growth rate, the industry, the location, everything we want that is a deal breaker or not, scoring everything and packaging it as a skill.
Second, we have the sourcing and enrichment agent. Kind of like what we have for marketing within the company. We have it for deal flow. So we can check the website, the tech stack, the recent news on the business, anything we want on the business. Third, this is really a game changer. We have an autonomous deal monitor. So every day we can get new listings on all the marketplaces out there with the broker email. You can pull new deals from the owner or the broker. It's super super cool. This is the prompt and I'll show you the output of that as well. Before this agent, we had to have like two analysts spending their full week just sourcing deals.
Then we also have a deal analysis and offer structuring agent. I'll put the prompt here as well. Most buyers send the same LOI template. We are able to scale it with the exact offer, our exact like green flags to make sure that the EBITDA and BSCR ratio and synergy model and perform all those models that I've shown in a different video. And this we also have the funding agent. Sort of like a bank and equity investor type agent. Let's say if you found a deal, you analyze it, you structure the offer for it, and the seller accepted. Now you need to fund it and this is where most deals die. So we can really, really help with that. We can introduce to banks, to investors. I'll show you the prompt that we helped with that to even help with the pitch deck for investors and the output from that.
Let me zoom out just quickly for a second. We have phase one is buying the business. We walked through all the agents, right? It's the deal flows, the analysis, the offers, the funding. Then we install AI in the deal itself and we prepare for exit. That's the future of M&A. AI will find deals for you, analyze them, structure them, raise capital or help you in the process and install AI to double the margins, help you position for exit and basically have the entire life cycle. Everything I just showed you, all 10 categories. All of that I'm going to teach in my next workshop. In the buy AI and sell workshop, I walk through every agent live. I'll show you how to install them in the business. I think it's right now on workshop.acquisitions.com. acquisitions.com. Sign up. It will be super super cool. If you want to work with us and get it done. If you just want to watch and see how we're going to do it, you can see it in the workshop for free. If you want us to do it for you, then book a call. I think it's dfy.acquisitions.com. Book a call with us and we can do that work actually for you to help you buy a business and install AI in your existing business or the business we help you buy. And um I think that's pretty much it. Hopefully this was valuable for you. I'm looking forward to get in touch with some of you and work with you and partner with you and invest alongside you or with you or in your deals. And let me know in the comments below what you think if you're still watching this and what would you like to see next. Thank you and I'll see you soon. Take care.